D postmortem

[bearophile]

Walter Bright:
> Can you look at the asm output and identify a case where it is less efficient?

You are right, I have cleaned my code up a bit, removing too much silly experiments. This is just a first version, if you need something different I'll try to show something different:
http://rafb.net/p/TDxobC64.html
Its asm:
http://rafb.net/p/16MemX81.html
The timings at the end are done on a dual core at 2 GHz with DMD V.1.029 (I can't use the two successive versions).

I think efficiency losses come from many different sources, so it may be difficult to spot a single small thing to be fixed.
(But DMD is a young language, at the beginning Java too was really inefficient, so for now I agree that it's better to focus most work on developing the language. Efficiency issues are mostly for later).

Thank you,
bearophile